Performance Engineering for a Medical Imaging Application on the Intel\n Xeon Phi Accelerator

We examine the Xeon Phi, which is based on Intel's Many Integrated Cores\narchitecture, for its suitability to run the FDK algorithm--the most commonly\nused algorithm to perform the 3D image reconstruction in cone-beam computed\ntomography. We study the challenges of efficiently parallelizing the\napplication and means to enable sensible data sharing between threads despite\nthe lack of a shared last level cache. Apart from parallelization, SIMD\nvectorization is critical for good performance on the Xeon Phi; we perform\nvarious micro-benchmarks to investigate the platform's new set of vector\ninstructions and put a special emphasis on the newly introduced vector gather\ncapability. We refine a previous performance model for the application and\nadapt it for the Xeon Phi to validate the performance of our optimized\nhand-written assembly implementation, as well as the performance of several\ndifferent auto-vectorization approaches.\n